Tramp patch for emacs-24 branch

Tramp patch for emacs-24 branch

[Michael Albinus]

Hi,

Last days, I have worked on a fix for a nasty Tramp problem: it loads
itself recursively, when default-directory is already a remote file
name. One could reproduce it with

emacs -Q -batch --eval '(let ((default-directory "/ssh::")) (expand-file-name "foo"))'

There is no related bug report; I've been pointed to this on #emacs of irc.freenode.net.

I believe I've fixed this, including a respective test in tramp-tests.el
for checks. But I'm wondering, whether the patch shall go into the
emacs-24 branch. It changes the autoload behaviour of Tramp, and cannot
be regarded as simple. And I believe the problem doesn't happen with
"normal" use of Emacs (whatever this is :-)

Comments? To be applied in emacs-24 or trunk?

Best regards, Michael.

Re: Tramp patch for emacs-24 branch

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Date: Wed, 16 Apr 2014 09:05:58 +0200
> 
> Last days, I have worked on a fix for a nasty Tramp problem: it loads
> itself recursively, when default-directory is already a remote file
> name.

And why is that a problem?  Does the second load cause any trouble?

> There is no related bug report; I've been pointed to this on #emacs of irc.freenode.net.

Did they say what did this cause them, apart of the extra loading.

> I believe I've fixed this, including a respective test in tramp-tests.el
> for checks. But I'm wondering, whether the patch shall go into the
> emacs-24 branch. It changes the autoload behaviour of Tramp, and cannot
> be regarded as simple. And I believe the problem doesn't happen with
> "normal" use of Emacs (whatever this is :-)
> 
> Comments? To be applied in emacs-24 or trunk?

I'd say it depends on how grave the problem is, see my questions above.

Re: Tramp patch for emacs-24 branch

[Michael Albinus]

Eli Zaretskii <eliz@gnu.org> writes:

Hi Eli,

>> Last days, I have worked on a fix for a nasty Tramp problem: it loads
>> itself recursively, when default-directory is already a remote file
>> name.
>
> And why is that a problem?  Does the second load cause any trouble?

It's not a second load, but an infinite recursive load:

emacs -Q -batch --eval '(let ((default-directory "/ssh::")) (expand-file-name "foo"))'
Recursive load: "/usr/local/share/emacs/24.3.50/lisp/net/tramp.elc", "/usr/local/share/emacs/24.3.50/lisp/net/tramp.elc", "/usr/local/share/emacs/24.3.50/lisp/net/tramp.elc", "/usr/local/share/emacs/24.3.50/lisp/net/tramp.elc", "/usr/local/share/emacs/24.3.50/lisp/net/tramp.elc"

> I'd say it depends on how grave the problem is, see my questions above.

It's a bug. But I believe the corresponding use case does not happen
often.

Best regards, Michael.

Re: Tramp patch for emacs-24 branch

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Cc: emacs-devel@gnu.org
> Date: Wed, 16 Apr 2014 09:34:21 +0200
> 
> >> Last days, I have worked on a fix for a nasty Tramp problem: it loads
> >> itself recursively, when default-directory is already a remote file
> >> name.
> >
> > And why is that a problem?  Does the second load cause any trouble?
> 
> It's not a second load, but an infinite recursive load:

Sounds pretty serious to me.

> > I'd say it depends on how grave the problem is, see my questions above.
> 
> It's a bug. But I believe the corresponding use case does not happen
> often.

But someone already bumped into it, so my vote is to apply to the
emacs-24 branch.  We've just had a first pretest, and there are still
a few serious problem we know about, so it's not like you will be in
danger of destabilizing a very stable code base.

Re: Tramp patch for emacs-24 branch

[Michael Albinus]

Eli Zaretskii <eliz@gnu.org> writes:

Hi Eli,

> But someone already bumped into it, so my vote is to apply to the
> emacs-24 branch.  We've just had a first pretest, and there are still
> a few serious problem we know about, so it's not like you will be in
> danger of destabilizing a very stable code base.

Thanks. I'll wait one day or two; if there's not a contrary opinion I'll
commit to emacs-24.

Best regards, Michael.